NC State
Hytönen, E., Sorsamäki, L., Kolehmainen, E., Sturm, M., and Weymarn, N. (2023). “Lyocell fibre production using NMMO – A simulation-based techno-economic analysis,” BioResources 18(3), 6384-6411.


The demand for man-made cellulosic fibres is expected to grow in the future. One commercially-available concept to supply fibres is Lyocell manufacturing from dissolving wood pulps using N-Methylmorpholine N-oxide (NMMO) as the solvent. The literature qualitatively indicates that NMMO recycling efficiency is a key factor for profitable operation. Process design information and parameter data are however poorly available publicly to illustrate the cost factors. Therefore, systematic techno-economic analysis of a 50 kt/year Lyocell plant was conducted using steady-state process simulation and cost modeling. With the simulation models, the underlying technical process design and modelling decisions, and economic assumptions were studied. NMMO makeup need is an important cost item. The simulated makeup need is very dependent on the design of the solvent recovery system and the vapor-liquid equilibrium thermodynamic model selection. On the other hand, water use, fibre washing process design, and washing model parameterization have relatively lower impact on the cost of production. Raw material cost and capital expenses are most critical cost items when the NMMO recycling efficiency is high.

Download PDF

Full Article

Lyocell Fibre Production Using NMMO – A Simulation-based Techno-Economic Analysis

Eemeli Hytönen,a,* Lotta Sorsamäki,b Erik Kolehmainen,a Michael Sturm,c and Niklas von Weymarn a

The demand for man-made cellulosic fibres is expected to grow in the future. One commercially-available concept to supply fibres is Lyocell manufacturing from dissolving wood pulps using N-Methylmorpholine N-oxide (NMMO) as the solvent. The literature qualitatively indicates that NMMO recycling efficiency is a key factor for profitable operation. Process design information and parameter data are however poorly available publicly to illustrate the cost factors. Therefore, systematic techno-economic analysis of a 50 kt/year Lyocell plant was conducted using steady-state process simulation and cost modeling. With the simulation models, the underlying technical process design and modelling decisions, and economic assumptions were studied. NMMO makeup need is an important cost item. The simulated makeup need is very dependent on the design of the solvent recovery system and the vapor-liquid equilibrium thermodynamic model selection. On the other hand, water use, fibre washing process design, and washing model parameterization have relatively lower impact on the cost of production. Raw material cost and capital expenses are most critical cost items when the NMMO recycling efficiency is high.

DOI: 10.15376/biores.18.3.6384-6411

Keywords: Lyocell; Textile fibre; NMMO; Modelling; Process simulation; Feasibility; Thermodynamic model

Contact information: a: Metsä Spring Oy, Revontulenpuisto 2 A, 02100 Espoo, Finland; b: VTT Technical Research Centre of Finland Ltd., P.O. Box 1603, 40101 Jyväskylä, Finland; c: Thüringisches Institut für Textil- und Kunststoff-Forschung Rudolstadt e. V., Breitscheidstraße 97, 07407 Rudolstadt, Germany;

* Corresponding author:


The process for producing Lyocell fibres from dissolving grade pulps using N-Methylmorpholine N-oxide (NMMO) as solvent is by now well developed. The development history of the amine oxide technology, called in this work NMMO-Lyocell process, until year 1999, has been reviewed e.g. by Perepelkin (2007) and White (2001). Several commercial NMMO-Lyocell production plants have been erected around the world during the last thirty years. In the 1990s, there were simultaneous developments in different countries. Courtaulds Fibres started three Tencel® plants, the first two in the USA and the third in the UK (Courtaulds part of Acordis) with original capacities of 16, 20, and 20 kt/a. Also, Lenzing built their first NMMO-Lyocell plant with capacity of 12 kt/a in Austria. Since then, Lenzing has expanded their production in Austria and acquired the plants originally operated by Courtaulds Fibres. It has quite recently also constructed a 100 kt/a plant in Thailand, which was started up in 2022. Recently, also other players in Asia and Turkey have initiated production and sales of Lyocell fibres. In search for sustainable alternatives for responding to the increasing need of textile fibres, other technologies using NMMO or other cellulose solvents have been developed (Golova et al. 2000; Sayyed et al. 2019; Makarov et al. 2020). In addition, modifications to the two slightly differing original NMMO processes by Courtaulds Fibres and Lenzing have been proposed (Wendler et al. 2012; Lidhure et al. 2019).

The NMMO-Lyocell process consists of two principal elements, the fibreline and the NMMO recovery. The first element is described in detail, e.g. in (White 2001), and the latter by Guo et al. (2021). From a plant design perspective, in addition to understanding the basic functions of the process elements, such as making dope, washing the fibre, or evaporating excess water from the dilute solvent, it is crucial to capture the reactions and phase changes of key process compounds. These phenomena have been studied experimentally: degradation and regeneration of NMMO (Rosenau et al. 2001), NMMO/H2O and NMMO/H2O /cellulose systems phase behavior under varying process conditions (temperature, pressure, composition) (Biganska and Navard 2003, 2004; Eckelt et al. 2009), and separation efficiencies of system compounds in fibreline functions, i.e. the mass transfer of key compounds (Hedlund et al. 2019). Further understanding on the phenomena is available outside the NMMO-Lyocell fibre production context. A noticeable body of scientific work has been created in the last 10 years in lignocellulosic biomass pre-treatment using NMMO. This has been summarized by Satari et al. (2019).

Systematic techno-economic analysis using process simulation and cost modelling plays an important role in assessment of process concepts and hence, in steering the development work. Using such bottom-up approach supports identification of the impacts of differences in scope and underlying assumptions related to the technology and economics of the considered concepts. Such analysis always encompasses several models that are used iteratively, and suitable modelling techniques and methods need to be selected carefully for each design stage. The goals and important limitations to be considered in an overall techno-economic assessment in different design stages are reviewed, e.g., by Hytönen and Stuart (2013) and Towler and Sinnott (2008). In any techno-economic study based on process simulation, capabilities of different methods for modelling of thermodynamics, physical properties, kinetics, individual process equipment, and process integration need to be evaluated and suitable methods selected. Moreover, in the economic assessment also the method for evaluating fixed costs and specifically for estimation of capital costs recalls consideration. A good analysis and guide for selection of capital cost estimation routine was published by Tsagkari et al. (2016). A key guiding principle, in selection of the models, is to assess first relevance of all models in the case and the availability of required input data for all considered concepts.

Very few studies present systematic process simulation of a Lyocell process, or economic assessment of processes using NMMO as the solvent. Bouwman (2008) used a computational fluid dynamics (CFD) approach to study certain process equipment in the fibreline of the Lyocell process. Shao et al. (2003) constructed a mathematical model of the spinning process to specifically simulate NMMO diffusion coefficient inside the fibre. In some studies, mathematical models have been used to describe certain phenomenon, such as NMMO/H2O system thermodynamics (Eckelt et al. 2009; Eckelt and Wolf 2008). The only plant-wide model of the complete process, reported in the literature to date, was developed by Reipsar (2020). He described the Lyocell process, and a model was used to simulate this process for studying different solvent recycling techniques. Modelling at a level higher than process modelling has been presented by Shen and Patel (2010). However, such models of the life cycle or supply chain are not commonly useful for process design, because they do not contain the process behavior in high detail. Specifically, no techno-economic assessments of Lyocell fibre production process have been published before. Only a few plant-wide process simulation -based process optimization studies of other fibre production processes have been reported (Bialik et al. 2020). Other context utilising NMMO as a lignocellulosic material pre-treatment solvent for production of ethanol or biogas from forest residues, pine or spruce, and food industry side-streams has been examined using process techno-economic analysis (Shafiei et al. 2011, 2014; Teghammar et al. 2014; Oliva et al. 2022).

In summary, based on the available literature, usable information for engineering process design and design analyses of Lyocell production process is very limited or non-existent. For example, overall techno-economic analysis studies of Lyocell processes are not publicly available. Furthermore, some papers discuss that fibre washing and NMMO recycling using evaporation are important costs defining process sections, but the needed process design parameter data and cost data are not available for verifying these claims. Therefore, the main goal of this work was to study the implications of the efficiency of evaporation and washing on the techno-economics of the NMMO-Lyocell production process. Process design and description of a feasible NMMO-Lyocell production plant, and a process simulation model and a cost model of this design, were developed to study these implications systematically. In the following chapters, the method, process design and models, and the analyses are described.


Overall Methodology

The overall method of the study is a steady-state simulation model-based techno-economic analysis. The used method is suitable for conceptual and pre-feasibility-level engineering analyses. In such analyses, the target process design and key design parameters are obtained from literature and other available sources, such as experimental work, expert opinion, or past projects if available. The design is then modelled using a simulation software to obtain the mass and energy balances. The models are used to describe key technical features and are the basis for economic analysis. In case when sufficient design and design parameter data are not available, engineering design knowhow is used to compile the needed data set to construct the models. Furthermore, developing the process design iteratively using process simulation, model parameter estimation can be conducted simultaneously.

The economic analysis combines price and cost data with the balance data to get variable production cost estimates in EUR/t. Capital cost estimates are obtained using methods suitable for concept screening or feasibility study (Christensen and Dysert 2005), and factorial methods are used to get fixed production cost estimates. Combining these, production costs can be obtained.

The next chapters describe in more detail the overall method used in this work. First, the hypothetical case study and the evaluated scenarios are described. Then the development of the process simulation model including the modelling of thermodynamics and fibre washing, as well as the approach for electricity consumption calculation are explained. Finally, the approach for conducting the economic analysis is presented.

Case Study

The scope of the hypothetical case study covers the fibreline of the Lyocell process (incl. pre-treatment, dissolution, fibre spinning, washing, and drying process steps) and the NMMO recovery and recycling process (incl. flotation/filtration, ion-exchange, evaporation, and NMMO regeneration process steps). Figure 1 presents the block-flow diagram of the process with all inputs and outputs. The process flowsheet is defined based on publicly available information, e.g. (White 2001; Zauner 2017; Jiang et al. 2020). The production of utilities (air, chemicals, cooling water, fresh water, and steam) and wastewater treatment are not included in the scope of the study. Instead, utilities are assumed to be available at the site, and they are considered in the economic analysis through cost contributions.

Fig. 1. Block-flow diagram of NMMO-Lyocell process. Abbreviations: CT = cooling tower, HEX = heat exchanger, NaOH = sodium hydroxide, HCl = hydrochloric acid, PG = propyl gallate, HA = Hydroxylamine, H2O2 = hydrogen peroxide, LP = low pressure. Line styles: solid olive green = fibre, double compound green = NMMO, dash-dot black = chemicals, dash yellow = LP steam, dash blue = steam condensate, solid light blue = process water, double compound-dash ice-blue = cooling water, solid red = air, solid grey = wastewater

In the pre-treatment block (pulper/press in Fig. 1), dry dissolving pulp is activated in a pulper by mixing with fresh (de-ionized) water. Sulfuric acid and sodium hydroxide can be used for pH-control. The activated pulp is pressed to as high dry matter as possible. In the dissolution block, the dewatered pulp is then mixed with recycled and makeup NMMO to form pre-dope. The pre-dope is mixed well in a ploughshare mixer, and water is evaporated under vacuum in a thin film evaporator to dissolve cellulose and form dope. Vacuum is used to prevent the degradation of NMMO solution which occurs if the solution is overheated. The evaporated vapors from dissolution are condensed and sent to NMMO recycling (evaporation block) to recover any evaporated NMMO. The dope is led to the fibre spinning block where it is drawn into a spinning bath. The bath is kept in constant temperature using cooling water, and in constant NMMO concentration by controlling the washing waters entering the bath. The bath receives continuously water from the washing block and the corresponding amount of bath liquid is led to NMMO recycling process. The formed fibre is led to a five-step counter-current washing block. Fresh water is used in the 5th washing step and recycled process water from solvent evaporation is fed to 4th washing step. The excess washing filtrate from each step is led to previous step and finally to spinning bath. Washing water amount is set to obtain the target residual NMMO content in the fibre, and to supply enough water to spinning bath to obtain the target NMMO content of the bath. The fibre is treated (finishing, cutting, bleaching) to targeted properties in post-treatment block. Washing water going to spinning bath is cooled using a cooling tower to reduce the need for cooling water. The washed and post-treated fibre is dried in the drying block using warm air. The drying air is circulated and mixed with fresh air before heating using low pressure steam. Circulation is controlled by adjusting air moisture close to dew point. Heat is further recovered from the exhaust air using heat exchanger.

The NMMO recovery process starts with flotation/filtration block, where solids and dissolved compounds are removed from the spent NMMO solvent. The partially cleaned solvent is then led to ion-exchange block where charged compounds are removed from the solvent. The anion and cation exchange columns are regenerated with sodium hydroxide and acid wash when their absorption capacity is reduced. The cleaned solvent is led to a six-stage co-current evaporation system. Pressure is decreased in the evaporators towards the last stage to keep the evaporation temperature in target range. Evaporation vents and the evaporated water from the last stage are collected, condensed, and fed back to the evaporation system. Water evaporated in three first stages is collected and recycled to fibre washing, whereas water evaporated in the 4th and 5th stage is recycled back to the evaporation system. NMMO-monohydrate is recovered from the last evaporation stage. NMMO degradation and regeneration is handled in the evaporation block, where the major part of the degradation product NMM (N-Methylmorpholine) is regenerated with hydrogen peroxide (H2O2) back to NMMO. Finally, the recovered NMMO-monohydrate is mixed with fresh NMMO-monohydrate makeup. Propyl gallate is added to reduce NMMO degradation. Hydroxylamine is also used to reduce fibre degradation.

The process design specifications for the considered NMMO-Lyocell plant are shown in Table 1. Even though several NMMO-Lyocell plants are commercially operating globally and thus the technology readiness level (TRL) is 9, actual plant and plant design data were not available for this study. Instead, all parameters are based on information available in the literature or expert opinions

Table 1. Key Process Design Specifications of the Considered NMMO-Lyocell Plant

Evaluated Scenarios

Four different scenarios were defined to study the implications of key process design specifications and modelling decisions. The scenarios are described in Table 2. To be able to draw distinct conclusions from the scenario comparisons, as many process design specifications as possible were kept constant.

The Base-case scenario describes the baseline of the process. The target of the Simplified recovery scenario is to reveal how the simplification of the NMMO recovery process (three-stage evaporation system and exclusion of NMMO regeneration) affects the economics. In Ideal thermodynamics scenario, the effect of the VLE model selection on the resulting economics is studied. The UNIFAC (universal functional activity coefficient) VLE model considers the molecular associations (interactions) between water and NMMO whereas Raoult’s law considers all compounds as individuals without any interactions. In the fourth scenario, the effect of reduced NMMO content in the spinning bath on the economics was in focus.

Table 2. Scenario Definitions

To enable transparent comparison, in the Ideal thermodynamics scenario three process design settings were adjusted: 1) the temperatures after evaporative process steps (dissolution, multi-effect evaporation in NMMO recovery) were set to the Base-case -scenario values by adjusting pressure profiles in evaporation, 2) water recycling rate was adjusted to reach the target fibre NMMO content, and 3) relative NMMO degradation was kept the same instead of targeting fixed NMMO recycling rate.

Process Simulation Model

Mass and energy balance model

For obtaining the mass and energy balances of the NMMO-Lyocell process, a steady-state simulation model is used. The process simulation model was developed using commercial steady-state simulation software Balas (Espoo, Finland).

The model topology is developed using MS Visio and the model is parametrized using the user interface of the software. The topology consists of units and streams; the streams transfer material information between the units that represent actual process equipment. The streams are in thermodynamic equilibrium in the conditions they leave their unit of origin, and they represent the process streams using three phases – solid, liquid and vapor.

Model compounds needed in the modelling of the NMMO-Lyocell process, and the phases they can exist in the model, are listed in Table 3. The thermodynamic properties of the solid-liquid model compounds are from Paccot (1987), and the properties of the liquid-vapor compounds are from Reid et al. (1977). The model compound named NMMO is a composite compound for all forms of NMMO in the system.

This simplification is needed to avoid managing many different model compounds in the model. The process is parametrized in such a way that the streams containing this compound are in molten state (the NMMO/H2O ratio and temperature above phase diagram line (Biganska and Navard 2003)).

Table 3. Model Compounds and Their Possible Phases in the Modelling Environment

User specified reactions, in addition to phase changes solved automatically by the simulator, are the dissolution of pulp-based compounds in dissolution block and the regeneration of dissolved compounds in the fibre spinning block. It is assumed that all pulp-based compounds dissolve completely, and no degradation of carbohydrate polymers to monomers occur. It is also assumed that 100% of dissolved pulp-based compounds will regenerate.

NMMO losses in the process are partly result of decomposition of NMMO to NMM and further to morpholine and formaldehyde. The major part of the degradation product (NMM) is regenerated with H2O2 back to NMMO (Kalt et al. 1999; Rosenau et al. 2001). Both the NMMO decomposition and regeneration reactions are considered in the model.

The simulation model is controlled using design constraints that can be met with certain set of design variable values. The design constraints fix the key process design parameters to targeted values. The simulator automatically solves the steady-state to fulfil the design constraints by varying the design variable values. The simulation software uses the SQP-solver (Sequential Quadratic Programming) to find the solution to the complex numerical problem. The constraints and variables are described in Table 4 for the Base-case.

In addition, several other model constraints (functions describing correlations between model parameters) are used. These are used for fixing other process design parameter values in cases where direct correlation exists (e.g., values given as %-value on dry pulp or fibre). These two sets of constraints enable flexible use of the model in sensitivity and scenario analyses.

Table 4. Design Constraints and Variables

Thermodynamic equilibrium modelling

Several thermodynamic models for describing the equilibrium state of liquid and vapor phase systems exist. In the process conditions of NMMO-Lyocell production, the liquid phase NMMO/H2O system can be described as a non-ideal solution, predominantly because of molecular associations (interactions) between the compounds. The vapor phase, on the other hand, can be assumed to behave as an ideal gas due to relatively low temperatures and water being the dominant compounds in vapor phase. To model the phase equilibria of such systems, activity coefficient -based methods and equation of states are useful (Mane and Shinde 2012).

In this work, the UNIFAC group contribution method was applied (Fredenslund et al. 1975, 1991). For fitting the group contribution parameters of the UNIFAC model, experimental VLE data for the NMMO/H2O system are needed. However, these data are not publicly available, and an expert opinion on the boiling point elevation of NMMO/H2O system was used instead as input data for deriving activity coefficients. For comparison, ideal thermodynamics would yield boiling point elevation of 18.5 °C for 86% NMMO/H2O system compared to the expert opinion of 45 °C.

NMMO/H2O/cellulose system (dope) is an even more complex system to model. The share of pulp-based organic compounds (cellulose, hemicellulose, lignin) in dope is 12.5 wt%. Phase diagrams of the ternary system NMMO/H2O/cellulose exist (Eckelt et al. 2009); however, they cover only solid/liquid equilibria. In this work, it was assumed that cellulose and other model compounds in solid and liquid phase do not influence the activity coefficients between water and NMMO, and that there is no activity between cellulose and other model compounds and NMMO and water. Cellulose and other model compounds are considered using the simulation software’s routines for ideal systems. Additionally, a scenario using simpler thermodynamic model was evaluated.

Washing modelling

Limited amount of performance data for washing Lyocell fibre is available in the public domain. On one hand, the model specification used in this work fully defines the inputs and outputs of the washing block: fixed NMMO content in the washed fibre after the counter-current washing and fixed NMMO content in the spinning bath can be met with certain fresh and recycled water flows. Such mass balance is possible to solve analytically. However, when the recycled water composition changes, as is the case in the considered scenarios, more detailed parameterization of washing is needed.

The Lyocell fibre washing system design assumed for the modelling follows the one described above and in literature (Zauner 2017). The same design of the washing block is used for all evaluated scenarios. The five washing stages are also assumed to behave identically. One convenient way to parameterize such washing sequence is to use three parameters for each washing step, as illustrated in Fig. 2. The three model parameters used to adjust the washing efficiency are i) dry content of the fibre exiting each washing stage, ii) shower wash water flowrate to each washing stage, and iii) K-value for NMMO in each washing stage. The K-value for NMMO (sorption indicator) is specified as mass fraction of NMMO in water phase in the outlet fibre stream per mass fraction of NMMO in water phase in the filtrate. K-value above 1 means that NMMO has sorption tendency to the fibres.

Fig. 2. Illustration of one washing stage in the washing model. The model parameters used to adjust the washing efficiency are highlighted. N = washing stage index, DC = dry content, = mass flow, filt. = filtrate.

In addition to these three washing unit parameters, the NMMO content of the fibre stream entering the first washing stage after the spinning influences the overall washing performance. Fibre NMMO content is dependent on the mass transfer of NMMO from fibre to the bath, which is driven by the bath NMMO concentration. The bath volume is assumed to be large in comparison to the fibre stream flowing through it, and the NMMO concentration is assumed to be constant in the bath. With the modelling level of detail suited for this work, concentration profiles inside unit operations (i.e. fibre inside the spinning bath) are not commonly considered. Instead, overall constant conditions are used to calculate output stream compositions in similar way as for washing (see above). Moreover, volumes (hold-ups) are not explicitly modelled in steady-state simulation. The effect of volume, or better described in this case the mass fraction of fibre (bath consistency in Table 1) in the bath at any given time, can be adjusted in the model using circulation of the bath content. The mass transfer in the spinning is assumed to behave similarly as in washing and therefore the same K-value is used.

To fix these four model parameters, sensitivity analysis was conducted for the Base-case scenario. The initial parameter values and the ranges used in the sensitivity analysis are shown in Table 5.

Table 5. Initial Washing Parameterization and Ranges Used in the Sensitivity Analysis of Base-case Scenario

Electricity consumption modelling

Estimation of electricity consumption without detailed process configuration, properties of process stream, and specific electricity demand of process equipment is challenging. Commonly, in early-stage conceptual studies, published electricity consumption values of the target process or similar process design contexts are adopted. More detailed analysis is however needed in this work to account for possible differences in the electricity demand in the considered scenarios.

Table 6. Assumptions for Electricity Consumption of Key Production Processes and Support Processes, Adopted from Pulp and Paper and Biobased Industrial Contexts. Weight Unit Refers to Input or Output of the Process Unit/Department